I have created a patch which can work with any audio file or a live input, but this patch is set up for a stereo environment. The left channel goes into 4 state variable filters: a low pass at 250Hz, a bandpass at 1k, a bandpass at 4k, and a high pass at 8k. The output of these filters goes into live.gain~ objects where the amplitude level at any moment is sent into a sub-patcher that slides the amplitude and then scales it to get to a delay time. Through early iterations, I found that while the sound file is playing, the majority of the file is spent between -30 and -10dB, so that is what I based the scale object on; however, also due to this I would get negative delay times, so I have it take the absolute value of the output of the scale object. This is then sent into a line~ object that smooths it over 50ms. The only difference between the left and right channels of audio is that for the left channel -30dB will give you a delay time of 10ms and -10dB will give you a delay of 1000ms. For the right channel, -30dB will give you 1000ms of delay while -10dB will give you 10ms.